Agentic AI and the SA cloud paradox

Organisations should approach AI spending in a sequential manner, building upon cloud modernization, rather than treating them as parallel tracks. Mark Walker, co-founder and director of Transformation for Innovation (T4i), emphasized this during his keynote at the ITWeb Cloud Summit 2026 in Riversands, Johannesburg. He noted that AI accounts for 20% of the South African enterprise CIO's IT budget in 2026, expected to grow to 50% by 2030, based on T4i's analysis.

Walker highlighted that security costs for AI in cloud environments increase to 10%-12% of total IT spend due to zero trust and edge protection requirements. He warned that underfunding security would lead to exposure and lack of control. The traditional lift-and-shift cloud migration era is now outdated in South Africa, replaced by a surge in hyperscale data centre investments and the rapid maturation of GenAI.

The local cloud market is projected to more than double, reaching over R100 billion by 2029. Walker pointed out a paradox for local enterprise leaders: while the infrastructure demands high-density, GPU-centric cloud environments for autonomous workloads, CFOs struggle with cloud bill shock and rand volatility from US dollar-denominated pricing.

Despite a marginal year-on-year growth of 0.8% in South Africa's ICT sector in 2025, driven by cloud and AI infrastructure demand, AI adoption remains limited. According to a Microsoft report, only 23.1% of South Africa's working-age population used AI in Q1 2026, up from 21.1% in 2025. This adoption rate places South Africa at 46th out of 147 economies ranked by AI diffusion.

Walker emphasized that productivity and efficiency are constraints to AI cloud execution, followed by a shortage of technical skills, data readiness, integration with legacy systems, and security concerns. He noted that South Africa's productivity rate is lower than competitors like China, India, Indonesia, and Vietnam. Walker identified micro-gains in developer velocity and basic customer-service automation, but also highlighted significant annual losses of R30 million per 1,000 employees due to operational friction.

Cloud and AI present a pathway to reclaim this value, he said. By 2027, operational efficiency gains are expected in back-office tasks such as finance, compliance, and legal document indexing. By 2031, AI could handle routine middle-management and clerical tasks fully. Walker stressed the importance of collaboration between CIOs and CFOs in driving an AI agenda, with each party owning critical aspects of the strategy.

The CIO's role involves bringing models to the data while keeping sensitive data in place and cutting unnecessary data egress. Modernizing core systems to be API-ready is essential for autonomous agents to query them effectively. Walker also stressed the need for an AI gateway to enforce identity, policy, and logging at the edge, and to eliminate shadow AI.

CFOs should prioritize non-negotiable FinOps through hard caps, tagging, and right-sizing before provisioning. Hedging currency exposure by contracting in rand and locking in longer terms where justified is crucial. Buying outcomes instead of seats by tying fees to measurable business value is another recommendation. Walker urged organizations to stabilize power and energy, secure compute access, redesign workflows around autonomous AI, mobilize the workforce, and fund rapid upskilling and talent attraction.

To address the shadow-AI gap, organizations should put governance structures in place ahead of AI policies.